4
$\begingroup$

Suppose $(S_n)_{n=1}^{\infty}$ is a simple random walk in $\mathbb{R}^2$. Let $\tau_{(a,b)}$ be the hitting of the first quadrant when $S_0 = (a,b)$. Is there a way to compute or estimate the distribution of $\tau_{(a,b)}$?

$\endgroup$
1
$\begingroup$

Asymptotic estimates for $P(\tau_{a,b}>n)$ as $n\to \infty$ can be obtained using the approach from http://arxiv.org/abs/1110.1254 (Random walks in cones). The results are valid for general random walks.

1) First we need to find the corresponding asymptotics for $\mathbf P(\tau^{BM}_{a,b}>t)$ for the Brownian motion. For that we can use paper Brownian motion in cones (doi:10.1007/s004400050111) which says that $P(\tau^{BM}_{a,b}>t)\sim \varkappa h(x,y)t^{-p}$ for a harmonic function $h(x,y)$ and some constant $p$. Harmonic function should be greater than $0$ on the continuation cone C and equal to 0 on the boundaries. Continuation cone $C$ in your question is a union of the 2nd, the 3rd and the 4th quadrant. For your question the required harmonic function will be $h(r,\phi) = r^{2/3}cos (\frac{2}{3}\phi-\frac{5\pi}{6}) $ in polar coordinates and p=1/3. Hence, $P(\tau^{BM}_{a,b}>t)\sim \varkappa h(x,y)t^{-1/3}$, as $t\to\infty$. Constant $\varkappa$ can be found in Corollary 1 of Brownian motion in cones (doi:10.1007/s004400050111)

2) Then you can use the main results of the above paper (Random walks in cones) to show that $P(\tau_{a,b}>n)\sim \varkappa V(a,b)n^{-1/3}, n\to\infty$ for the function $V(x,y)$ which is harmonic for the killed random walk. We know that $V(a,b)\sim h(a,b)$ when $a$ and $b$ are large and far away from the boundary of the cone $C$.

There is also a recent paper discussing your question http://arxiv.org/abs/1511.02111 This paper also gives asymptotics for $P(\tau_{0,0}>n)$, see Remark 2 on page 5.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.